Over the past few days, several replacement video figures have been seen on video numbers, resulting in the transformation of China’s explosive video into a foreign-dominated video, which is of great interest in some of the electric-business scenes, helping cross-border electricity owners to follow a multi-market video. Of course, the same is true of the transfer of overseas explosive funds to the country。
Image from video number screen
For electric operators, after a domestic video was taken, they did not want to find another foreign model to reshoot it, so that model replacement technology could be used。
Next, let me say a few technical ways to achieve this result, helping small partners in need to open their minds and find ways for themselves。
For many women, many grass videos are presented by models. For markets with similar skin colours, such as those in China and North America, it is possible to reduce costs by changing the face of a Chinese model to that of an overseas model because of the close colour。
2. Dancing models
there was a time when the dance video character was hot, and if we thought about it, we would find that the dance video character replacement could actually be derived from other areas, such as basketball, as long as there was no big cover-up, and we could pretend to be a dance video with a character action video and upload it to the model. the representative product is viggle, which not only replaces dancers, but also selects objects by editing template capabilities, for example, by replacing basketball in basketball videos with the basketball brand you want to promote。
3. Nano Banana + Act-Two
Runway has access to the Act-Two function, and it provides an ability to drive character movements with video, and technically and viggle dances are similar, but runway works better. Then why use Nano Banana? Runway's act-two has two drivers: first, the person in the drive picture moves by the person in the video, and second, the person in the video adjusts to the expression in the video. We're actually going to use the first one, and we're going to provide a picture, and runway will let the people in this picture do the same thing as the people in the video that you're sending. It doesn't sound like a replacement with the character in the video. But what if we provide the same picture as the first frame of the video, except the character? Yeah, that's why we use Nano Banana, because so far, only Nano Banana can be very consistent。
According to the illustrations above, not only can we replace people with this option, but the products in our hands are easy and easy so that the same shot can be used in different product presentations。
4. ComfyUI + Wanx-Vace
In the ComfyUI ecology, the open source of the popular Vace model is free of charge and becomes the most fully functional and capable video-editing model of the moment, which is used extensively in video role replacement. The quantitative version of the Vace model can also run on a consumer-grade graphic card, which can be enjoyed if it costs thousands of dollars to buy a medium- and high-end card。
Presentation by Network
The programme, which is used in the field of cross-border electricians, can address the objective of the replacement of couriers。
5. The violence mapping programme
Comfy's working hard? I am here to offer a simple, brutal and heinous solution. This approach, which we used more than 10 years ago, is not as efficient as the current AI tools. That's how it's gonna be
we all know that the video is made up of a frame, and in that case we replace it with a frame based on a specific sample rate, with a person-to-person replacement for each frame, and we can use the flux-kontext-dev and some of the plugins used to control the person's attitude, multi-wire batch run maps, and then spelled the picture into a video. such a programme would be easier to understand, but it would also be possible to map in depth certain frames, the disadvantage being that the cards would be scrap。
wrap-up
THE PAPER PROVIDES DETAILS OF FIVE ALTERNATIVES TO VIDEO CHARACTERS, WHICH, IN PRACTICE, PRIMARILY TAKE INTO ACCOUNT COST AND TIME. TECHNICALLY, VIDEO MODELS ALLOW FOR THE REPLACEMENT OF PEOPLE, MAINLY BY USING BODY-LED TECHNOLOGY, AND USER UPLOADS OF VIDEOS, MAINLY BY PROVIDING DRIVER-DRIVEN BONE INFORMATION FOR THE REPLACEMENT OF PEOPLE IN VIDEOS. AS AI MATURES, IT BECOMES EASIER AND EASIER TO DO IT, BUT IT GIVES THE REPLACEMENT ELEMENTS ON THE WEB LINE, AS ILLUSTRATED BELOW
Theoretically, this function can directly replace the characters in the video, but unfortunately, I've tried it several times, and it's not working。
When we are free to modify the video, the sound and make it more and more real, attention should also be paid to using those technologies in a compliance setting rather than using some illegal channels. As ordinary people, we should also be vigilant about different types of video。